#8 Mismatch in simple CMOS current mirrors - Part I

This video discusses the fundamentals of random current mismatch in MOS current mirrors. The underlying equations are explained with examples and concepts are explained to make intuitive sense. Mismatch variation in strong and weak inversion regions are explained with the help of plots.

[@malashis]

You explain everything very well. Simple, and to the point. Thanks.

[@asha503]

All ur videos are very very informative....Thanks for uploading such very informative , nicely explained ,in depth videos...which won't get much in books and all...very experienced designers only can explain this much depth concepts....it will be very useful for designers...thanks for sharing knowledge in depth...👍👍👍.... would like to see many more videos...

[@asha503]

Nicely explained 👍👍

[@shlooky1]

great lecture! thank you!

[@-.-aaa]

12:30 In weak inversion, mismatch for both cases should be the same since sqrt(WL) for both cases are the same? Not sure why you said the 2nd device has to be 4 times?

[@hemantshukla3098]

Quite informative. Many thanks for this. When we do simulation we have options of running Monte Carlo sim (time consuming) vs a DC match. Would you please tell in case of current mirror when DCmatch sim is useful and when to run a Monte Carlo? May I request you to please cover the simulation aspect of designing a current mirror for let’s say a desired mismatch no. And how to optimise for Flickr noise and mismatch spec.

[@analogsnippets]

That's a good point. I think DC mismatch is a good simulation for a quick estimate but Monte Carlo should be the sign off. So on final design Monte Carlo should be run. Noise is closely related to mismatch but a different topic on its own. So perhaps deserves a dedicated Snippet. Something to work on in future. Thanks for the comments!

[@hemantshukla3098]

Analog Snippets can we assume a DC match sim useful for local mismatch around a fixed DC point whereas Monte Carlo incorporates a larger variation. Yes I agree for quick calculations DC match is fast and quick for evolving design

[@analogsnippets]

Yes that's correct. Monte Carlo is like large signal analysis while DC mismatch is like small signal analysis around DC operating point. DC mismatch gives quick estimates based on probability calculations.

[@abhiruplahiri1]

What is a DC match simulation? I have always used MC and never run a DC match before.

[@analogsnippets]

It is a new type of simulation which gives a quick estimate of mismatch. Not a traditional simulation.

[@98505177229850590818]

while doing monte carlo simulations, how would you observe this current mismatch? my thought is you can do this by taking the current quantity ratios from two mirror FETs as extract parameter and then doing monte carlo on those parameters for statistical device models ? is that right. ?

[@analogsnippets]

Don't quite understand the question. Monte Carlo can be done in a DC simulation by properly biasing the transistor and plotting/measuring the drain current. If you want to see current variation in one transistor only then you can bias this transistor at a desired current and run a DC Monte Carlo. This biasing can be done using a negative feedback around transistor using ideal circuit elements (high gain vcvs). If you want to see current variations in a current mirror then it can be biased using ideal current source.

[@98505177229850590818]

If possible please make videos on random vs systematic vs process mismatch differences

[@analogsnippets]

It's good suggestion. Thanks. Will add to the list.